To See Myself, I Wrote 99 Stories

If I asked you "Who are you?", you might tell me your name. If I asked again, you might tell me your gender, your profession, your hometown, your hobbies. But what if I asked a third time? Most people would probably call me crazy and walk away, but perhaps some would begin to seriously discuss who they really are.

You may not have explicitly asked yourself "Who am I?", but you've probably been curious about it. Some people, to satisfy this curiosity, turn to astrology and fortune-telling to understand their past and future. Others take psychological tests, trying to distill personality into symbols.

These are all good approaches—I'm somewhat familiar with them myself. But after experiencing them firsthand, I always felt their answers were bland, unable to satisfy me. So I thought, maybe this time I should do something serious about it.

The Feeling of Dissatisfaction

I've always believed that many so-called "tests" don't really provide a direct answer to "who you are"—they just create a hazy, mysterious atmosphere and the "surprise factor" when you see the results. They might come from some ancient mystical civilization, or just require answering seemingly profound questions. They appear so magical, as if completing them would immediately grant you insight into your inner world and find life's answers and direction.

MBTI

The most successful in this regard is MBTI—the Myers-Briggs Type Indicator. This is a personality test consisting of three scales designed to measure different innate preferences in how the brain perceives and makes decisions. These three measurement tools interpret you from three angles: innate traits, environmental factors, and future development.

If you've taken what you thought was the MBTI, you might find it strange that you never encountered three separate scales—you just kept clicking through one big questionnaire and got an answer. Here's where it gets interesting: what you actually took wasn't the MBTI but something called 16Personalities. This scale is essentially a modified Big Five personality test that forcibly maps its conclusions onto the MBTI system's classification labels.

Why would they do this? I suspect it has to do with product packaging. The Big Five personality test has been around for years, but we don't hear about people taking it for fun because it lacks that "surprise factor"—and this "surprise" is an essential element for internet products to open wallets and generate buzz.

This surprise factor is built across many dimensions. One aspect is catchy labels that tell you what kind of person you are, as if slapping it on your forehead would eliminate all confusion for life. Before, you might say "I'm naturally introverted" or "that person is crazy," but now all descriptors have become I-types and E-types. You no longer need to retrieve adjectives from your brain—just know sixteen letters.

In contrast, Big Five personality test results are incredibly complex—major dimensions containing sub-dimensions, each with scores that can be interpreted differently based on high or low values. Reading through everything would probably put you to sleep, naturally lacking the refreshing snap of binary labeling.

Another aspect is exquisite packaging. You'll rarely encounter such refined visual experiences in legitimate psychological tests—beautiful illustrations, consistent outstanding style, and psychologically-flavored interpretations. This "professional, international" feel adds to its credibility.

Finally, there's SEO engineering. If you Google "MBTI," 16Personalities—which has almost nothing to do with actual MBTI—will definitely appear on the first page, adorned with hooks like "you can test for free," fishing for users, even catching psychology professionals without high psychometric scores. Mind you, the "genuine" MBTI is a commercial psychological assessment product that costs money.

Precisely because MBTI testing has been confused to the point where even professionals fall into the trap, I maintain considerable reservations about popular science resources and marketing activities centered on MBTI: Which MBTI are you discussing? Do you actually know which MBTI you're talking about?

If you ask me, I definitely prefer the Big Five's interpretation style. Even if 16Personalities has proven its excellent reliability and validity through research, I still don't think its balance between commercial and professional considerations is appropriate.

SCL-90

Another widely misused tool is the SCL-90 (Symptom Checklist-90). Friend, let me tell you—those of us who study psychology regularly receive requests like this: "My XX friend is in terrible psychological shape," accompanied by an SCL-90 for me to analyze. While my facial expression remains calm, internally I'm already having a "fishwife's tantrum" because using such scales rashly without professional counselor guidance and crudely interpreting their results often brings negative effects.

Many people over-interpret scores, viewing "high scores" as "having mental illness," but SCL-90 is only a screening tool—it cannot be used for diagnosis. High scores merely reflect current stress states and emotional fluctuations, not mental disorders.

Some ignore the test's temporal validity. The scale assesses symptoms from "the past week," but they make long-term judgments about themselves based on single test results. Mental states change dynamically over time—we can't use momentary observations to make biased interpretations of a person's long-term traits.

Oversimplifying complex problems is also common. "Unpracticed" players easily reduce complex psychological states to a few fluctuating numbers, thinking high scores mean "psychological problems" and low scores mean "fine," ignoring individual differences and various situational factors.

Most dangerous is self-labeling—giving yourself labels based on high scores in certain dimensions, like "I have OCD" or "I have depression." These symptoms might be temporary or mild, but once labels are applied, they can have profound effects.

If you only learn one thing from this article, I believe this is most important: Almost all clinically-used scales need to be used under professional counselor guidance and interpreted by professional counselors. Self-testing with scales found online is universally dangerous.

Marketing-Purpose Psychological Tests

As for those tests that certain red-icon apps regularly produce for entertainment, claiming backing from professional papers while dressing themselves up attractively—I can only say, you do you [smile].

Dissatisfied with What?

I want to better understand myself, but even after studying psychometrics, I haven't seen any handy tools. Even if I took every questionnaire available, I'd probably just end up with a dozen reports and a pile of labels that still couldn't answer the opening question:

Who are you?

To answer a question, you need to define it. The need before me was too abstract and vague, so I began trying to define it precisely, constraining the question's boundaries to make it a manageable project.

My master's research included considerable content related to brain mechanisms in intimate relationships, with attachment patterns as a core topic—how early life experiences influence social behavior. What interested me most was our emotional responses when facing external stimuli.

High Permission

If we were to describe what you and I share impartially, I think "organism" is a good word, especially that character for "life." Survival is our instinct—nearly every function in our brains revolves around higher survival rates. "Emotions" play a crucial role in this.

Emotions have been functioning since our birth. Human infants are among the most helpless young in nature. They cannot move independently, forage, or defend against predators. This prolonged, extreme vulnerability constitutes one of the most severe survival challenges in human evolution. Any trait that could improve infant survival rates would be preserved and strengthened under natural selection pressure.

Take crying—it's an infant's most powerful weapon. Crying is a high-decibel signal that triggers intense physiological and psychological discomfort in adults (especially mothers). This discomfort drives caregivers to take immediate action, checking the infant's needs: Is the child hungry, in pain, cold, or needing comfort? Infants whose cries better attracted attention were more likely to receive timely care, thus improving survival rates.

Or consider smiles—if crying is "demanding," then smiling is "rewarding." An infant's social smile is a powerful "adhesive." When caregivers meet an infant's needs, a smile or giggle provides enormous emotional reward and satisfaction to exhausted caregivers. The underlying brain mechanism involves releasing neurotransmitters like oxytocin and dopamine. These releases create positive feelings that encourage continued caregiving behavior, strengthening parent-child bonds.

As infants receive increasingly rich external stimuli, they continuously learn what things will harm them and what stimuli benefit survival.

Fire is a good example. Infants initially might not instinctively fear dancing flames—on the contrary, their light and heat might arouse curiosity. But when they try to touch and feel burning pain, or when they see caregivers' tense expressions and warnings about fire, the brain rapidly establishes a powerful negative association: fire is dangerous, I fear fire.

Food provides an excellent example of positive emotions. When a child first tastes sweet berries, taste buds activate, and a delighted smile spreads across their face. This positive emotional experience forms positive feedback in the brain: sweetness equals pleasure, pleasure equals behavior worth repeating. Thus, children become more inclined to seek similar foods, reinforcing their preference for "deliciousness." This mechanism not only helps us identify high-energy foods but also keeps us open to exploring new foods—an important survival advantage in resource-scarce ancient environments.

Of course, not just fire and food can activate emotions. From colors and touch to people and environments, when emotions occur, various cognitive elements present might be remembered by our brains and trigger corresponding emotional responses when similar elements appear.

Any mechanism involving survival must operate extremely quickly, especially when facing potential threats or opportunities—this automatic response can determine life or death in an instant. If you, currently reading this passage, suddenly heard a loud noise with room walls violently shaking, the amygdala in your brain would rapidly activate, triggering a "fight or flight" response. Heart rate accelerates, muscles tense, adrenaline surges—these physiological changes occur within less than a second, even before we consciously analyze "what is this," the body has already prepared to respond to danger.

Imagine if in such situations we had to wait for the prefrontal cortex to slowly extract concepts, clarify logic, weigh pros and cons, and make decisions—we'd already be buried under rubble.

This rapid emotional response has its costs. When emotional intensity is too high, the brain's attention becomes completely occupied, consciousness "bandwidth" narrows, and our rational thinking abilities may be temporarily suppressed. Excessive anger might cause people to lose reason and act impulsively; extreme fear might freeze someone in place, unable to act. This is our inner "animal side." In this state, deep brain structures temporarily "take over" the brain, suppressing the prefrontal cortex's rational regulatory functions, immersing us completely in these emotional experiences.

High Function

Fortunately, we don't live constantly at risk of being buried under rubble. Most of the time, emotional triggers aren't immediate life-or-death threats, giving space for "wisdom's brilliance" to shine. In fact, our prefrontal cortex can intervene and regulate the entire emotional development path, assessing current situations and adjusting or redirecting our cognition to display humanity's "civilized side."

This regulation can take many forms. "Attention redirection" is a common strategy. When feeling angry due to work frustrations, you might stand up to gaze at scenery outside the window or put on headphones to listen to favorite music—shifting attention away from the source of negative emotions, thereby weakening emotional intensity.

"Cognitive reappraisal" is also common. When you feel angry and depressed about test scores below expectations, leading to self-abandonment and thinking yourself "incompetent," some people might immediately "hit the brakes," stop negative emotions from running wild, and "reappraise" their cognition as an opportunity to "discover knowledge blind spots and accumulate experience for future success." This can guide us to solve problems—if emotions stem from specific difficulties, the rational brain begins analyzing problems and planning steps, transforming emotional energy into problem-solving motivation. Even in certain social situations, it blocks impulses to speak harshly or display inappropriate emotions, maintaining surface interpersonal harmony and dignity.

Our prefrontal cortex tags, categorizes, and interprets these emotional experiences based on their occurrence across people, events, time, places, and objects, giving us quite rich emotional vocabulary. Accurately identifying emotions is important—it could even be said this is the foundation of all advanced emotional regulation strategies.


If we cannot distinguish whether what we experience is "disappointment" from unmet expectations or "anger" from perceived injustice, it's like a doctor unable to diagnose what illness a patient has—naturally, there's no way to prescribe appropriate treatment. Someone with vague perception of their emotional states might lump all negative feelings under "stressed" or "in a bad mood," thus only able to adopt the crudest coping methods like overeating or entertainment addiction. Or when facing others' negative emotions, only able to say devastatingly low-EQ thunder like "don't be so unhappy," instantly "exploding the scene."

Homeostasis

After understanding surface logic, let's look deeper.

High school biology textbooks once taught that biologically, our bodies are homeostatic systems. Homeostasis refers to biological organisms maintaining relatively stable internal environments when external conditions change through self-regulatory mechanisms. For example, when external temperature drops, our bodies generate heat through muscle shivering and reduce heat loss through blood vessel constriction to maintain constant body temperature. When blood sugar concentration is too high, insulin secretion increases, promoting glucose entry into cells, controlling blood sugar within normal ranges.

However, our psychological level also has a homeostatic mechanism. But this stability isn't the unchanging, serene kind of "stability"—it's a dynamic balance built on our brain's continuous prediction of the future. Extensive neuroscience research proves our brains constantly make predictions about upcoming events based on past experience and current information.

If everything that happens in the future fits perfectly within expected scripts, then everything is peaceful—we won't have overly intense emotional reactions. Like waking up in the morning and following established plans to wash, eat breakfast, and go to work, with each link unfolding as expected, our inner world maintains a calm flowing state. But once this stable state is broken, we generate a "need" to return to stability, triggering corresponding emotional reactions.

What breaks homeostasis might be external stimuli. For example, rain suddenly starts pouring outside, and you realize the bedding you expected to dry nicely is about to get soaked—naturally, you become nervous and rush outside to save tonight's sleep. Or you're working intently when suddenly you receive a message saying an important meeting has been moved up, disrupting your orderly plans and instantly flooding you with anxiety. These are external "expectation violations" triggering our emotional reactions.

Homeostatic disruption can also come internally. For example, your stomach suddenly rumbles, you remember you haven't eaten lunch, hunger breaks bodily balance, generating a "need to fill your stomach." Or when alone at home, you suddenly think of a long-out-of-touch friend, loneliness quietly strikes, generating a "need for social connection."

Of course, not all homeostatic disruptions trigger negative emotions. Positive "expectation violations" similarly activate our emotional systems. You thought it was just an ordinary day but unexpectedly received acceptance for a long-desired job; or you randomly bought a lottery ticket that won a small prize; or you accidentally found an eye-opening good book in a bookstore. These positive events exceeding expectations also break our psychological homeostasis, but bring surprise, excitement, and satisfaction, pushing us to share this joy or continue exploring more possibilities.

So our logic chain here isn't as simple as previously described. A more accurate description is: internal and external stimuli break our "expectation homeostasis," the brain detects this deviation, we generate pressure to return to homeostasis—that is, "need"—the emotional system begins working, and finally the prefrontal cortex interprets and regulates it.

Taking Out the Mirror

Stumbling Start

Reading this far, it's clear we have two completely different systems in our brains: a "high permission" system managing immediate emotional responses, and a "high function" system managing advanced cognition. The "high permission" system can rapidly take over the body in emergencies, triggering instinctive responses to threats; while the "high function" system uses prefrontal rational analysis to make more "socialized" decisions based on complex reality conditions. "Need" is the driving force that makes the entire system start turning. When reality conflicts with our expectations, homeostasis breaks, and a "need" emerges.

In my view, understanding oneself means understanding these three systems' internal structures and collaboration methods. How are they activated? How do they influence each other? These internal structures largely define who we are, why our behavioral patterns and logic are what they are, and how we respond to life's ups and downs.

So how do we understand this system's operation? As someone with formal training, my first thought was to "make questionnaires"—create a Likert five-point scale¹, measure all emotional associations in pairs, and directly get a network graph that would somewhat reflect the emotional operating logic of the test-taker (myself).

I was then contemplating how to define "two emotions being associated" and how to conduct statistics and analysis. But, holy shit! My friend! This was a thoroughly terrible idea! If you've ever bought a "Contemporary Student Survival Manual," you'd know how rich emotional vocabulary expression is—my version had 361 words. If I wanted to brute-force scan everything, I'd have to do 129,960 multiple choice questions. I'm not a computer—I can't just run two for-loops and call it done! ( TдT)!

Then I fell into a "user experience" death spiral, mentally constructing various methods to try reducing question quantity and response difficulty. Like changing Likert five-point assessments to binary "yes/no" judgments; or cutting the entire assessment matrix into different zones and distributing them on circular diagrams for connection, providing a new dimension in operation and visualization to accelerate assessment processes and reduce cognitive load.

But no matter how you put it, 130,000 questions is too ridiculous. Just doing dozens of questions is already painful—the most perverted Minnesota Multiphasic Personality Inventory only has 566 questions. When I was in university, my counseling psychology teacher once recalled the horrific scene when the teacher had the whole class take that scale—nearly everyone went numb.

But what other approaches were there? Even lying in bed unable to sleep at night and racking my brains, I couldn't think of a suitable path.

Playing Big!

It happened that during this period, I wanted to work with a friend on an emotion-awareness-centered project. We talked about holes I'd dug but not filled when writing the survival manual.

At the time, the survival manual planned a companion website with emotional concepts distributed according to emotion wheel logic. We designed an elegant interactive experience to guide users through exploring the entire emotional space, where users would collect these emotions and record current events with brief text.

This project never executed due to some engineering flaws and several unclear product points, so it sat on the hard drive collecting dust. I'd occasionally pull out the whole idea and chew on it again, but unfortunately, every time I couldn't taste anything new.

Later we also made an emotion card deck—each emotion was a card, each card had pictures. I tried using this deck to design emotion understanding and awareness tools, but that also went nowhere.

But ADHD is like this—sometimes you unexpectedly knead seemingly unrelated things into one whole lump. One day I suddenly realized I could reconsider these several dusty projects and unsolvable problems within one system. "Emotional narrative" might be a viable solution. Then my heart sprouted an impulse to do something, so I opened Typst, shrunk all the cards and re-typeset them into stickers, printed three large A4 sheets of adhesive downstairs.

My brain's idea was that since "emotional words" are learned through continuously experiencing external stimuli, seeing vocabulary should help us find the most fitting memories. So I opened my browser and typed out several life stories. After finishing, I took out the small notebook I usually use for practicing calligraphy, opened unused page backs, and transcribed them.


Actually, transcription was redundant and time-consuming, stickers were also redundant and expensive to print, but I felt these stories deserved recording in a more formal way, deserved adorning in a more exquisite format, contained in a carrier less likely to disappear, so I dove in head first. I wrote a story and stuck the corresponding small sticker on it—wrote sixty pages total.

The so-called stories weren't the kind with background setup, characters, rising action, climax, and conclusion all laid out in ink, but shallow three or four sentences accurately recording people, events, time, places, objects and my feelings. I wrote in groups of three words—one positive, one negative, and one need.

Considering writing over three hundred sixty stories was too perverted, I reduced the scope to the emotion wheel's first three layers, totaling one hundred thirty emotional words. Since pairing was needed, the number of writable stories depended on whichever group had the fewest words, and also depended on my expressive ability, willingness, and past experiences. So initial planning was to write 99 stories.

Ah ha! This time I'm playing big! So I began some bizarre "diligent writing." Really nearly sparked fire from grinding my pen (　ﾟ 3ﾟ).

Initial writing was light and pleasant—after all, who can't sit there pondering past events? So it was pure output. But when I reached the 60th story, my thinking began gradually exhausting. On one hand, this process itself was part of product research—naturally I hoped to quickly obtain results, so I pushed myself rather hard, initially writing six stories daily, gradually increasing to maybe eighteen daily. Since it simultaneously involved multiple different emotions, and while writing I'd also be immersed in those experiences, I was quite exhausted, even feeling very gloomy and bitter inside for several days. On the other hand, I really couldn't think of stories to fill these spaces.

But things gradually turned around. As I shifted attention to different life periods, some dusty memories began slowly emerging. Many memories were quite pleasant, relaxed, and joyful, so I didn't stop at the 99th story—my brain still had content to continue writing. The final submission was 105 stories, with even the last few stories being very important "closing statements."

These stories covered all aspects of my life, from early childhood to entering the workplace, from what I like eating and watching to sexual harassment, attacks, and bullying I experienced, from past brilliant successes to unspeakable failures. I felt simultaneously like a doctor and a patient, calmly facing my naked self, seriously observing and even appreciating every detail.

I should provide some examples here, but considering my own shame levels and readers' reading experience, I'll allow myself to keep particularly dark examples private. Here I'll only share some relatively general-audience ones, like:

Positive emotions:

Attraction: I remember it was a summer when I was still a student. I woke up very early for some reason—the sky was already half-bright, showing a very tranquil gray-blue color. The lawn downstairs had just been mowed, emitting a refreshing fragrance. Since my family lived on the top floor, I could clearly see that complete sky. I was deeply attracted watching this peaceful scene, thinking I still had to go to school so I quickly lay back down, but then got up again to look for a while. That's an image I'll never forget in this life.

Negative emotions:

Shame: I must admit there's a very dark side to my personality. In high school, my relationship with my desk-mate was quite good—his grades were exceptionally excellent and he was good at sports, while in contrast I was almost his opposite due to ADHD and physical weakness. I remember once during Chinese poetry recitation testing, he passed completely while I didn't. I lost my temper, he pointed out the ugly jealousy in my heart, and my homeroom teacher also patiently helped me see this point. Though I was quite unconvinced at the time, thinking about it now, I'm truly ashamed.

Needs:

Consistency: I really like "orderly" things, so I always buy notebooks of the same brand and size, ink from the same brand, even take pills out to put in same-sized stainless steel bottles, put seasonings in same-sized jars. Even when designing, I'm unusually obsessed with "alignment," because this sense of order makes me feel refreshed and delighted.

I started this experiment on June 30th, completed the 99th story on July 9th, and finished six additional stories on July 10th.

Playing with Data

I think 99 stories is a good standard. On one hand, it provides enough pressure that I couldn't escape facing some real emotions; on the other hand, it provides some space without being too demanding that I mechanically fill spaces with unimportant stories just to "complete the task." From a data analysis perspective, taking only n=30 to "sample" a complex "person" really seems inadequate, even somewhat argumentative and irritating in my view.

My planned analysis included two parts: qualitative analysis and quantitative analysis. Qualitative analysis provides an overview report, objectively and calmly telling me what kind of person I actually am. Quantitative analysis provides a clear structure, allowing me to understand my reaction patterns when facing different people, events, times, places, and objects in a structured way.

Qualitative Analysis

Though called qualitative analysis, I still hoped the entire structure could have basic reproducibility, so I needed a basic report structure and execution standards. To clarify this content, I separately chatted with Gemini, Claude, Grok, and DeepSeek—four different model versions—about my ideas several times. I first gave the entire process a proper name: E99 Emotional Narrative Assessment. Of course, this isn't any strictly rigorous psychological assessment—I just hoped to have clear targets and execution direction during the entire discussion process.

Each model gave me very different but inspiring answers, so I extracted discussion fragments that met my needs and mashed them into a mixed, fragmented requirements document, giving it to "Master of Laws" for final integration. Honestly, I must admit that if this kind of requirements document came from a real product manager, they'd probably be dragged out and hanged. By this standard, I who wrote such an unsightly document am probably also unforgivably evil! But I'm facing an emotionless LLM—what's there to be ashamed or worried about! (chest out).

After such organizing, plus my final manual adjustments, I got this report architecture:

• Introduction (Core Summary and Key Insights): At the report's opening, use highly condensed text to summarize this person's core portrait.
• Foundation and Background (Source of "Who Am I"): This part aims to explore the person's "growth soil," explaining the origins of their personality, values, and trauma.
• Internal Structure and Dynamics (Mechanism of "How I Operate"): This is the report's core, depicting a "living" person. It shows how the "foundation" shaped in part one operates, conflicts, and manifests in real life.
• Development Trajectory and Direction (Outlook on "Where Am I Going"): Based on analysis from the first two parts, this section assesses the person's future direction and clarifies their current core driving forces.

Of course, the entire document has many details, like specific paragraph structures for each part, how to develop and discuss, what to do and not do. Due to space limitations, I can't paste the complete prompt here, but the above structure should help you understand the final report's general appearance.

I ran different models several times. Though occasionally there were off-track situations, under prompt constraints everyone wrote basically similar content, differing only in writing style, but core qualitative analysis work had no major problems.

Quantitative Analysis

Quantitative analysis was relatively more complex. On one hand, you can't have LLMs do Likert scales—this behavior itself reeks of unreliability. On the other hand, we still had to solve the massive pairing problem. Even with only 105 final story samples, there were 10,920 pairs needing comparison—even for large language models, this is too cruel.

After much thought, my final approach was having LLMs find the 200 most associated story pairs and explain what their associations were.

To ensure "association" definition was clear, I referenced methods used in qualitative analysis, organizing a "What Constitutes Association" document and clearly defining four types of associations:

• Thematic Similarity: Sharing themes or situations, for identifying core life domains;
• Causal or Temporal Sequence: Temporal or logical causal links, for tracking emotional evolution;
• Psychodynamic: Similar internal conflicts or defenses, exploring psychological structures;
• Character Relationships: Same significant individuals, for analyzing relationship dynamics;
• Key Element Overlap: Shared people, events, objects, or places, highlighting influential life elements.

But cutting it down this way, something still seemed off—we still had to answer questions similar to "significance": "Why do you say there's association just because you say so?" As interest research, I gave a relatively violent answer: repeatedly run the same task 10 times, collecting association frequencies as "certainty basis." Personally, I felt 10 times might not be sufficient, but after running this quantity, some patterns were beginning to emerge, so I temporarily stopped tormenting the models, integrated all data, and tried presenting them visually.

Since the entire requirement wasn't very complex, I made a small website—importing data to final result presentation took less than an afternoon. The finished product looks like this:


What you see before you—this brain-like image—is my emotional network. This network's final shape is completely determined by algorithms and data; I have no control over it. So honestly, seeing this brain-shaped three-dimensional figure after completion, I felt somewhat shocked inside, especially since my master's specialized in neuroscience. From this shape, I could still feel some "mysterious fate."

Node colors represent node types: red for negative, blue for positive, brown for needs. Lines between nodes indicate how many assessments pointed out associations between these nodes.

Clicking middle connections shows which assessments indicated connections between them, what types of connections they are, and the reasoning large language models gave for connections. Following this network all the way through, I found that for very thick connections, underlying reasoning was quite consistent. This is good.

After extracting network data, I hoped to see some patterns from this network, so I decided to tag nodes. The approach was similar—feeding raw data to LLMs, requiring extraction of people, events, time, places, objects from each story, making tags and writing them into raw data. To ensure tags were comprehensive, this task was repeated 12 times.

Compared to classification, tagging is a relatively open task, so output might have some randomness, making output data cleaning relatively critical. I used strict data processing here:

First, expand tag data. For each tag of each node, check if other tags are contained within the current tag—if so, add the contained tag to that node's tag set. For example, if one node has "homeroom teacher" while another node contains "middle school homeroom teacher," then "middle school homeroom teacher" would "split" out a "homeroom teacher" tag.

Then clean the data:

• Length filtering: Tag length must be greater than 1 character, filtering out single-character tags;
• Emotional word filtering: Tags cannot duplicate existing emotional words;
• Node frequency filtering: Tags must appear at least 4 times in a single node, ensuring tag importance;
• Node distribution filtering: Tags must appear in at least 3 different nodes, ensuring tag universality.

This entire cleaning process aims to: remove noise tags, avoid functional confusion, ensure tags are both sufficiently important in single nodes and have cross-node universal significance, improve tag data quality and usability in network graphs.

With this node tag data, we can play some new tricks. The image below shows node filters—if you click a tag, all unrelated nodes and connections disappear, showing only relevant information.


The image's right side shows two networks: an "ADHD" network and a "high school" network, relatively objectively reflecting the emotional landscape of ADHD and high school life in my eyes.

These "sub-networks" look very much like "constellations." When introducing these sub-networks to others, I also like using the word "constellation" to greatly reduce introduction length, because this word itself contains the implication of "making inferences about personality." But unlike constellations, the "stars" and "patterns" here have clear meanings and can be evidenced.

Speaking of constellations, I must mention Mr. Zeng's famous joke: "You believe people have free will, but also believe people born in the same month all have the same personality?" Mentioning this isn't to prove anything, just that you've already read nearly ten thousand words and might feel bored—I wanted to lighten the mood σ`∀´).

What's in the Mirror?

While writing this article, I kept clicking around this network graph, trying to observe and understand the information from different angles. Seeing some content confirmed my guesses, while clicking on other information felt fresh—turns out these two experiences have such connections. I sometimes zoom out of the network, sometimes drill inside to look up at those edges. Though seemingly chaotic, they present structure and beauty. I feel like I'm in a magnificent building, gazing at humanity's complexity and wonder.


I'm also thinking—I seem fortunate to be in this era, because with such advanced technology, I have the opportunity to transform learned knowledge and conceived ideas into a visible form. How magical, how wonderful.

Facing that real report and the complex network, my heart is filled with mixed emotions.

Though not a peer-reviewed test, I still presumptuously named it E99, in some sense hoping that people might put aside misuse of SCL-90 and obsession with those labels. One letter is more forceful than three letters, and 99 is bigger than 90, conforming to traditional aesthetic recognition that "bigger is better." I know this is some kind of spiritual victory method—this self-made toy will never serve any genuinely substantial purpose. But through this personally polished mirror, I indeed see different scenery. For me personally, this experiment's significance and impact are already sufficiently profound.

If you had asked me "Who are you?" before, I might have given this answer:

Two eyes, one nose, one mouth.

But what do I see from the mirror now? I think Gemini summarized it well:

He picks up fragments of reason from broken past, weaving with creativity as thread a well-ordered island to resist the world's invasion. He is both this island's hard guardian and a prisoner yearning for the mainland's warmth. That highly sensitive heart intoxicates in moments of beauty and aches in silent alienation. His life's journey is searching on this lonely sea for a harbor where he can remove his armor and anchor safely.

That's today's sharing, thanks for reading through.